GRAID NuImages: Spatial Reasoning Questions for Vision-Language Models

GRAID NuImages is a question-answer dataset generated by the GRAID framework for enhancing spatial reasoning in Vision-Language Models. The dataset was created by author kd7 and is associated with a research paper and project page. It was last updated on October 29, 2025.

Use Cases

• Training VLMs on spatial reasoning based on generated question-answer pairs.
• Benchmarking VLM performance on spatial understanding tasks.
• Developing new methods for high-fidelity data generation for computer vision.

Strengths

• Dataset is associated with a published research paper and dedicated project page.
• Framework (GRAID) is specifically designed for creating spatial reasoning datasets from object detection annotations.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

Generated from object detection annotations, likely from the NuImages dataset.

Collection Method

Created using the GRAID (Generating Reasoning questions from Analysis of Images via Discriminative AI) framework.

Freshness

Last updated 2025-10-29 01:02:33; freshness should be verified.